Sequential product on effect logics

Transcription

1 Sequential product on effect logics Bas Westerbaan Thesis for the Master s Examination Mathematics at the Radboud University Nijmegen, supervised by prof. dr. B.P.F. Jacobs with second reader drs. J. Mandemaker, written by Bastiaan E. Westerbaan, student number , on August 15th Second corrected version, September 26th 2013.

2 Abstract In categorical logic predicates on an object X are traditionally represented as subobjects. Jacobs proposes [9] an alternative in which the predicates on X are maps p: X X + X with [id, id] p = id. If the coproduct of the category is well-behaved, the predicates form an effect algebra. So this approach is called effect logic. In the three prime examples of effect logics, a sequential effect algebra arises naturally. These structures are studied by Greechie and others in quantum logic. In this thesis we study several variations on effect logics, and prove that in these variations sequential effect algebras do not arise. 2

3 1 Introduction 1.1 Starting point Given a category C with coproducts and an object X of C. Following [9], define the (internal) predicates on X as ipred(x) = {p: X X + X; [id, id] p = id}. In Set, the category of sets, the predicates on X correspond to subsets: ipred Set (X) = { { x x U p U ; p U (x) = κ 2 x x / U ; U X}. In Kl(D), the Kleisli category of the distribution monad, the predicates on X correspond to maps X [0, 1]: ipred Kl(D) (X) = { p ψ ; p ψ (x) = ψ(x) x + (1 ψ(x))κ 2 x; ψ : X [0, 1] }. In Hilb, the category of Hilbert spaces with (bounded linear) operators, the predicates on X correspond to operators on X: ipred Hilb (X) = { p A ; p A (x) = (Ax, x Ax); A: X X }. If the category C is well-behaved, then ipred(x) carries an algebraic structure: it is an effect module. We will cover effect modules and related structures in Section 2. An effect module has (among other structure) a partially defined binary operation, a unary operation ( ) and a selected element 1. In the previous examples: C p q defined whenever p 1 Set U V U V = X U X Kl(D) ψ + χ ψ + χ 1 1 ψ 1 The predicates ipred(x) on a Hilbert space X do not form an effect module. However, the predicates that correspond to operators 0 A I, called positive predicates ppred(x), do form an effect module. C p q defined whenever p 1 Hilb A + B A + B I I A I In Set and Kl(D), there is an obvious way to extend the map X ipred(x) to a functor ipred: C EMod op. We write (f) for ipred(f). In the case of Hilbert spaces, the map X ppred(x) can be extended to a functor on the wide subcategory of Hilbert spaces with isometries: ppred: Hilb isom EMod op. We have three functors: Set Kl(D) Hilb isom ipred ipred ppred EMod op EMod op EMod op Assume we have any functor Pred: C EMod op. Any effect module is also a partially ordered set for each : X X + Y we have an order preserving 3

4 map ( ) : Pred(X +Y ) Pred(X). In our three examples, the map ( ) has a left and right order adjoint: ( ). Also, for each p Pred(X) there is an evident map char p : X X + X such that (char p ) 1 = p. On any Pred(X), we define p? (q), pronounced p andthen q, by In our examples, we have: p? (q) = (char p ) q. C Set Kl(D) Hilb isom p? (q) U V ψ χ AB A These three operations are examples of Sequential Products as defined by Gudder and Greechie [7] to study Quantum Logic. In fact, these are their prime examples as well. This leads to the following question, which is the starting point of this thesis: are there categorical axioms, which our examples obey, such that andthen is a sequential product as defined in [7]? 1.2 Overview First, we will cover in Section 2 the basic theory of several algebraic structures related to effect modules. Also a few specialized results will be proven, for instance on the existence of certain effect monoids, which will be used in the study of effect logics later on. Then, in Subsection 2.4, we introduce the sequential product as defined by Gudder and Greechie in [7]. We conclude the preliminaries by recalling some basic topics, such as galois connections, monads and the Kleisli category. We will assume the reader is familiar with Hilbert spaces and C -algebras. For an introductory text see [2]. To a lesser extend, we will assume familiarity with Category Theory. For an introductory text, see [1]. Before we will investigate effect logics in the line of [9], we will investigate, in Section 3, a weaker notion of effect logic, called weak effect logic. In a weak effect logic we start with a functor Pred: C EMod op of which we do not require that that Pred(X) = ipred(x). First we look at several examples of weak effect logics. Then, we prove a representation theorem to characterize the possible andthen that occur in a weak effect logic. In Section 4, we study effect logics. First we study internal predicates. Then we introduce the axioms of an effect logic and consider some examples. After that, we prove three representation theorems to partially characterize the andthen that occur in an effect logic. Finally, in Section 5, we summarize our results and state the open problems. We introduce some notions and prove some theorems that are not directly required for the results of this thesis. These are marked by a. We include these for one of two reasons. Either it is a negative result to justify our approach. For instance, the non-commutative effect monoid we construct in Subsection is rather complicated. One might expect that there is a finite example. That is 4

5 why we include Proposition 40, that states every finite effect monoid is commutative. Or: the result is a worthwhile deviation. For instance, to show there is a non-commutative effect monoid (Corollary 51) we only require one direction of the equivalence between OAU-algebras and convex effect monoids. However, the full result is worth proving. 5

6 Contents 1 Introduction Starting point Overview Preliminaries Effect algebras Some results on infima and suprema Isotropic index Interval effect algebras Convex effect algebras Lattice effect algebras Finite effect algebras Effect monoids Convex effect monoids and OAU-algebras Effect monoids on finite dimensional lexicographically ordered vector spaces Effect modules Sequential effect modules Examples Counterexamples Galois connections Kleisli category The distribution monad for an effect monoid Coproducts and split monos Weak effect logics Examples Basic theory Representation of weak sequential effect modules Effect logics Internal predicates Axioms Examples Set Kl(D M ) Hilb Representation theorems Conclusions Summary Further investigation Acknowledgments

7 2 Preliminaries Before we can fully state the axioms of (weak) effect logics, we need to introduce some algebraic structures of which the effect algebra, effect module and effect monoid are the most important. Also, we will use the notion of a Galois connection (or order adjunction). One class of effect logics is based on the Kleisli category of a distribution monad generalized to an arbitrary effect monoid. This class of effect logics is used to prove Theorem 116. Due to this theorem, we will study examples of non-commutative effect monoids. 2.1 Effect algebras In an effect logic, the set of predicates on an object will have (among others) the following algebraic structure. Definition 1 ([3]). Given a structure E,, 0, 1 where : E E E is a partial binary operation: we write a b whenever a b is defined and 0, 1 E are selected elements: 0 is called the zero and 1 the unit. This structure is called an effect algebra (EA) if the following holds. (E1) (partial commutativity) If a b then b a and a b = b a. (E2) (partial associativity) If a b and a b c, then b c, a b c and a (b c) = (a b) c. (E3) (unique orthocomplement) For every a E there exists a unique a such that a a = 1. (E4) If a 1, then a = 0. Given two effect algebras E and F a map f : E F is an effect algebra homomorphism if 1. f is additive, that is: when a b for a, b E, then also f(a) f(b) and furthermore: f(a) f(b) = f(a b) and 2. f preserves the unit, that is: f(1) = 1. The effect algebras along with their homomorphisms form a category called EA. Example 2. The following are effect algebras. 1. [0, 1], +, 0, 1 where [0, 1] R is the unit interval and + is the normal addition. x y whenever x + y 1 and x = 1 x. 2. P(X),,, X where P(X) is the set of subsets of X and is the disjoint union: A B whenever A B =. The orthocomplement is the normal complement: A = X A. 3. Eff(H ), +, 0, I where Eff(H ) are the positive operators on a Hilbert space below or equal to I; I is the unit operator and + is addition of operators. A B whenever A + B I and A = I A. 7

8 There is more structure on an effect algebra: there is a partial order and a difference. Before we define these, we need to derive some basic properties. Proposition 3. In any effect algebra, we have 1. (involution) a = a; 2. 1 = 0 and 0 = 1; 3. (zero) a 0 and a 0 = a; 4. (positivity) if a b = 0 then a = 0 and b = 0 and 5. (cancellation) if a b = a c then b = c. Proof. One at a time. 1. By (E3), we have a a = 1. Then by (E1), we have a a = 1. And thus by (E3) again, we must have a = a. 2. By (E3), we have 1 1 = 1. Then by (E4), we have 1 = 0. By the previous 1 = 1 = By (E3) and (E1), we have a a = 1. By the previous (a a) 0 = 1 0 = 1. Thus by (E2) we know a (a 0) = 1. Hence by (E3) we have a 0 = a. 4. By (E3) and the previous, we have (a b) 1 = 0 1 = 1. Then by (E2) we have b 1. Hence by (E4) we know b = 0. And similarly, using (E1), we see a = Since (a b) (a c) = (a b) (a b) = (((a b) a) b = 1, we know by uniqueness of the orthocomplement that b = ((a b) a) = c. Thus: effect algebras are partial commutative monoids with the extra axioms (E3) and (E4). Definition 4. We write a b if there exists a c such that a c = b. Proposition 5. For any effect algebra E. 1. E, is a poset; 2. a b if and only if b a ; 3. 0 is the minimum and 1 is the maximum element; 4. if a b and b c, then a c and a c b c and 5. a b if and only if a b. Proof. One by one. 1. First, reflexivity: a 0 = a thus a a. Then, anti-symmetry: suppose a b and b a. That is: there are c, d E such that a c = b and b d = a. Then a 0 = a = b d = (a c) d = a (c d). Hence by cancellation c d = 0. Thus by positivity c = d = 0. Consequently a = b d = b 0 = b. Finally, transitivity: suppose a b and b c. Then there are d, e E such that a d = b and b e = c. Hence c = b e = (a d) e = a (d e). Thus a c. 8

9 2. Suppose a b. Then a c = b for some c. Note that (b c) a = b b = 1. Thus b c = a. Hence b a. Conversely, suppose b a. Then by the previous a = a b = b, as desired a = a hence 0 a. Thus 0 is the minimum. In particular 0 a. Then by the previous: a = a 0 = 1. Thus 1 is the maximum. 4. Suppose a b and b c. There is a d such that a d = b. Hence a c d = a d c = b c. Thus a c b c. 5. Suppose a b. Then (a b) a b = 1. Hence by uniqueness of orthocomplement: (a b) a = b. And thus a b. Conversely, suppose a c = b. Then b b = a c and thus by (E3) and (E2) in particular a b. Note that if a c = b = a c, then by cancellation c = c. Definition 6. Suppose a b, let b a be the unique element such that we have a (b a) = b. Proposition 7. For any effect algebra E. (D1) x y is defined if and only if y x; (D2) x y x; (D3) x (x y) = y and (D4) if x y z, then z y z x and (z x) (z y) = y x. Proof. One by one. (D1) By definition. (D2) x (x y) = x hence x y x. (D3) (x (x y)) (x y) = x = (x y) y and thus, by cancellation, we derive x (x y) = y. (D4) x (y x) (z y) = y (z y) = z and hence by uniqueness of the difference we have (y x) (z y) = z x and thus z y z x. ((z x) (z y)) z x = ((z x) (z y)) (z y) y x = (z x) y x = z y and thus by cancellation ((z x) (z y)) x = y and finally by uniqueness of the difference: y x = (z x) (z y). Remark 8. Any X,,, 1 in which (D1) (D4) hold and has 1 as largest element is called a D-poset. Thus any effect algebra is a D-poset. Conversely: on any D-poset we can define a b = c c b = a and a = 1 a. This will form an effect algebra. Thus: D-posets are another way to look at effect algebras. 9

10 Proposition 9. For any effect algebra E, we have 1. a = 1 a; 2. if b a then (a b) = a b; 3. if b, c a and a b = a c, then b = c; 4. if a b, c and b a = c a, then b = c; 5. if a b, then (a b) b = a and 6. a (b c) = (a b) c whenever they are both defined. Proof. One at a time. 1. (1 a) a = 1 and thus the unique orthocomplement a = 1 a. 2. Certainly (a b) b = a. Thus (a b) b a = 1. Hence by uniqueness of the orthocomplement we know (a b) = a b. 3. Suppose a b = a c. Then by the previous: a b = (a b) = (a c) = a c and thus by cancellation b = c. 4. Suppose b a = c a. Then b = (b a) a = (c a) a = c. 5. Certainly b a b and ((a b) b) b = a b. Cancelling: (a b) b = a. 6. Certainly a b b. Thus a b b. Hence (a b) c b c = (b c). Thus (a b) c b c. Consequently (a b) c (b c) = (a b) b = a. By uniqueness of the difference (a b) c = a (b c). Definition 10. Given a b, define a b = (a b ). Proposition 11. For any effect algebra E. 1. (a b) = a b ; 2. (a b) = a b ; 3. a b = a b and 4. E = E,, 1, 0 is as an effect algebra isomorphic to E. Proof. One by one. 1. (a b) = (a b ) = a b 2. (a b) = (a b ) = a b 3. By the previous: a b = (a b) = a b = a b. 4. The map x x is its own inverse and thus a bijection. By the previous, 0 = 1 and 1 = 0 the effect algebra it induces is precisely E. Hence x x is an isomorphism between E and E. By definition, we only require an effect algebra homomorphism to preserve and 1. This is enough for it to preserve the other structure as well 10

11 Proposition 12. For any additive f : E F, we have the following. 1. (order preserving) If x y, then f(x) f(y). 2. Whenever y x is defined, we have: f(y x) = f(y) f(x). 3. (preservers zero) f(0) = 0 If additionally, f preserves the unit (and thus f is an effect algebra homomorphism), then also the following holds. 4. (preserves orthocomplement) f(x ) = f(x) 5. If x y, then f(x y) = f(x) f(y). Proof. One at a time. 1. Suppose x y. Then there is a c such that x c = y. Thus f(x) f(c) = f(x c) = f(y). Hence f(x) f(y), as desired. 2. Suppose y x is defined. By definition y = (y x) x. Thus f(y) = f((y x) x) = f(y x) f(x). By uniqueness of the difference: f(y) f(x) = f(y x). 3. Certainly 0 = 0 0 and thus 0 f(0) = f(0) = f(0 0) = f(0) f(0). Hence by cancelling: 0 = f(0), as desired. 4. By definition 1 = x x. Thus 1 = f(1) = f(x x ) = f(x) f(x ). By uniqueness of the orthocomplement, we know f(x ) = f(x). 5. By definition x y = (x y ). Thus by the previous f(x y) = f((x y ) ) = (f(x) f(y) ) = f(x) f(y) Some results on infima and suprema We will look at the order structure of an effect algebra in more detail. The results we prove, will be useful when we consider effect algebras that are lattice ordered in Subsection If we consider the partial binary relations, and with one argument fixed and restricted to its domain, we see they are either order isomorphisms or order antiisomorphisms. Proposition 13. Given an effect algebra E. 1. b a b is an order isomorphism from a to a. Its inverse is the map b b a, an order isomorphism from a to a. 2. b a b is an order isomorphism from a to a. Its inverse is the map b b a, an order isomorphism from a to a. 3. b a b is an order antiisomorphism from a to a, which is its own inverse. Proof. We already saw that the maps are appropriately order preserving or order reversing. Also we saw that the cancellation law holds for all these operations, hence all maps are injective. It is left to show that each map is defined on the given domain; maps into the given codomain and is surjective. 11

12 1. a b is defined whenever b a. Thus a is indeed the domain. Furthermore a b a. Given any a c. Then a (a c) = a. Hence the map is surjective. This also show that b a b is its inverse. 2. Note that b a = (b a ) = b a. Thus the previous pair of maps with a for a are exactly the current maps. Hence these are also order isomorphisms. 3. a b is defined whenever b a. Thus a is indeed its domain. Furthermore a b a. By (D3) the map is its own inverse. Thus in particular, it is surjective. A very useful corollary of the previous is that the operations, with some restriction due to their partial definition, either preserve or invert suprema and infima. Corollary 14. Given an effect algebra E and a subset U E. Write a U = {a u; u U}. And similarly for the other operations. Write a U whenever a u for each u U. 1. Suppose U a, then a U exists if and only if a U exists and we have a U = a U. Also a U exists if and only if a U exists and we have a U = a U. 2. Suppose U a, then a U exists if and only if a U exists and we have a U = a U. Also a U exists if and only if a U exists and we have a U = a U. Proof. Note that if U a, then also U, U a, whenever they exist. Thus the suprema and infima in the order restricted to a, are the same as in the whole of E. Similarly for a. The first part is now an easy consequence of the fact that a ( ) is an order isomorphism, from a to a, which preserves suprema and infima. The second part is similar. One of the applications is the following proposition. Proposition 15. Given an effect algebra E and a, b E. If a b and a b exists, then a b = (a b) (a b). Proof. Certainly a, b a b. And thus by the previous corollary, we have (a b) (a b) = ((a b) a) ((a b) b) = a b. Corollary 16. The previous proposition has some useful consequences. 1. If a b = 0 and a b, then a b = a b. 2. Whenever it is all defined: (a b) (a b) = a b Isotropic index In this section we introduce terminology that we will use when we study finite effect algebras in Subsection and lexicographically ordered vector spaces in Subsection Definition 17. Given an effect algebra E. 12

13 1. An element e is called isotropic if e e. 2. Given n N and an e E. We can define 0e = 0 and (n + 1)e = ne e whenever ne e. That is: ne is e summed n times with itself. 3. If ne is defined, but (n + 1)e is not, then n is the isotropic index of e; in symbols: ord(e) = n. 4. If ne is defined for all n N, then e is called infinitesimal and we write ord(e) =. 5. If a is infinitesimal and for all n N we have na b, then a is infinitely smaller than b and we write a b. 6. If 0 is the only infinitesimal, then we call E Archimedean Interval effect algebras The effect algebras [0, 1] and Eff(H ) we saw before are examples of a more general class of effect algebras: those that are derived from partially ordered abelian groups. Definition 18. A structure G, +,,, 0 is called a (partially) ordered abelian group provided 1. G, +,, 0 is an abelian group; 2. G, is a partial order and 3. if a b then a + c b + c for any a, b, c G. An element a G of a partially ordered group is called positive if 0 a. We write G + = {g; 0 g} for the positive elements. Given two elements a b in an ordered group G, we define the (order) interval with endpoints a and b as [a, b] = {c; c G; a c b}. Example 19. The following are examples of ordered abelian groups. 1. R, +,,, 0, the real line with addition. 2. B(H ) R, +,,, 0, the Hermitean operators on a Hilbert H space where the order is defined as follows. A B if Ax, x Bx, x for all x H. Proposition 20. Given any ordered group G and strictly positive element 0 < u, the structure [0, u], +, 0, u is an effect algebra with a = u a. Such an effect interval is called an interval effect algebra. Proof. Assume 0 a, b, c u. (E1) If a + b u, then a + b = b + a u, as desired. (E2) If (a + b) + c u, then (a + b) + c = a + (b + c) u, as desired. 13

14 (E3) 0 a thus u u + a hence u a u + a a = u. thus 0 = a a u a. Thus u a is in [0, u]. Also a u Clearly a + (u a) = u, thus u a is an orthocomplement of a. Given any other b [0, u] such that a + b = u. Then b = u a. Thus the orthocomplement is unique. (E4) Suppose a + u u. Then 0 a 0. Thus a = Convex effect algebras When we consider interval effect algebras derived from ordered vector spaces, the effect algebra inherits a scalar multiplication from the vector space. For a few proofs it is useful to introduce the notion of a scalar multiplication on any effect algebra. This definition will turn out to be equivalent to that of a [0, 1]-effect module. Definition 21. An effect algebra E is called convex if for every λ [0, 1] and a E there exists a λ a such that (C1) α (β a) = (αβ) a; (C2) if α + β 1, then αa βa and (α + β) a = α a β a; (C3) if λ [0, 1] and a b then λ a λ b and λ a λ b = λ (a b) and (C4) 1a = a. However, we did not discover anything new: any convex effect algebra is an interval effect algebra of some ordered vector space. Definition 22. Given an ordered vector space V over R and a vector u > 0. We say [0, u] generates V if for every v V there are r 1, r 2 R and v 1, v 2 [0, u] such that v = r 1 v 1 r 2 v 2. Theorem 23 (S. Gudder and S. Pulmannová [5]). For any convex effect algebra E, there exists a unique ordered vector space V and an u > 0 such that [0, u] = E and [0, u] generates V Lattice effect algebras Another class of effect algebras are those derived from orthomodular lattices. Definition 24. A lattice is a partial order for which each finite supremum and infimum exists. A bounded lattice is a lattice with a maximum element 1 and a minimum 0. An orthocomplemented lattice is a bounded lattice together with a unary operation ( ) such that 1. (complement) a a = 1 and a a = 0; 2. (involution) a = a and 3. (order-reversing) if a b then b a. An orthomodular lattice L is a orthocomplemented lattice such that for any a, b L, we have: if a b then a (a b) = b. 14

15 Example 25. The following are orthomodular lattices. 1. Any Boolean algebra is an orthomodular lattice with as orthocomplement the normal complement. In particular P(X),,,, X, X ( ). 2. Given a Hilbert space, the partial order of its closed linear subspaces by inclusion is an orthomodular lattice. We can extend any orthomodular lattice to an effect algebra. prove this, we need a lemma. Before we Lemma 26. In any orthocomplemented lattice, the laws of de Morgan are valid. That is: we have (a b) = a b and (a b) = a b. Proof. a b b thus (a b) b. Also (a b) a. By the definition of infimum (a b) a b. For the other inequality, we first note that clearly a b a. Thus a (a b ). Also b (a b ). Hence a b (a b ). That is (a b) a b. We proved (a b) = a b. The proof of the other equality is dual. Proposition 27. Given any orthomodular lattice L,,, 0, 1, ( ). Define a b if a b and then a b = a b. The structure L,, 0, 1 is an effect algebra. Furthermore, the order of the effect algebra is the same as the order on L. Proof. To prove that L,, 0, 1 is an effect algebra. (E1) Suppose a b. Then a b hence b a, since ( ) is order reversing. Thus b a and a b = a b = b a = b a. (E2) Suppose a b and a b c. Then a b and a b = a b c. Certainly b a b c. Thus b c. Also c (a b) = a b. Thus c b (a b ) c = a, by the orthomodularity since b a. Hence a b c. Thus we are justified to write: (a b) c = a b c = a (b c). (E3) a a = a thus a a hence a a = a a = 1. Indeed: a is an orthocomplement. Suppose a b = 1 for some b. Then b a and a b = 1. Thus by orthomodularity a = b (b a ) = b (a b) = b 1 = b 0 = b the orthocomplement is unique. (E4) Suppose a 1. Then a 1 = 0. Thus a = 0, as desired. Now we prove that the order of the effect algebra on L is the same as the order of the lattice. Suppose that there is a c such that a c = b. By definition, we have b = a c = a c a. Conversely, suppose a b. Then by orthomodularity we have a (a b) = b. Certainly a a b, thus a a b. Hence a (a b) = b. We saw that every orthomodular lattice can be extended to an effect algebra. This extension is, in fact, unique. Definition 28. If the order of an effect algebra is a lattice; that is: finite infima and suprema exist; then it is called a lattice effect algebra. If the order of an effect algebra is an orthomodular lattice, then it is called an orthomodular effect algebra. 15

16 Proposition 29. If E is an orthomodular effect algebra, then a b = a b. Proof. Given a b. Then a b. Certainly b 1. Thus by modularity b b = b (b 1) = 1. Hence 0 = 1 = (b b ) = b b. Consequently a b b b = 0. By Corollary 16, we see a b = (a b) (a b) = (a b) 0 = a b Finite effect algebras Another obvious class to investigate are the finite effect algebras. Definition 30. An effect algebra E is called finite if it has a finite number of elements. Definition 31. Given an effect algebra E. An element a E is called an atom if 0 < a and for every b < a we know b = 0. Proposition 32. Given a finite effect algebra E such that 0 1 and a 1,..., a n are its atoms. Then for each e E there exist e 1,..., e n N such that e = e 1 a 1 e n a n. Proof. First we prove that for every e > 0, there exists an atom a such that 0 < a e. If e itself is an atom, we are done. If not, there must exist an e 1 such that 0 < e 1 < e. Now we consider e 1. If it is atom, we are done. If not, there must exist an e 2 such that 0 < e 2 < e 1. And so forth. Since E is finite, there cannot exist an infinite strictly decreasing sequence 0 <... < e 3 < e 2 < e 1 < e. Thus there must be an atom a < e. Now we prove e is a sum of atoms. If e is an atom or equal 0, we are done. If not, there exists an atom a 1 such that 0 < a 1 < e. Then 0 < e a 1 < e. If e a 1 is an atom, we are done. If not: there exists an atom a 2 such that 0 < a 2 < e a 1. Then 0 < e a 1 a 2 < e a 1 < e. If e a 1 a 2 is an atom, we are done. If not: then we repeat with e a 1 a 2. This procedure must end, for otherwise we would find an infinite strictly decreasing sequence. Definition 33. Given a finite effect algebra E with atoms a 1,..., a n. A tuple (t 1,..., t n ) N n is called a multiplicity vector if t 1 a 1 t n a n = 1. Let T (E) denote the set of multiplicity vectors. The multiplicity vectors determine a finite effect algebra. Definition 34. Define T = {a; a N n ; t T (E). a i t i for all i}. For a, b T, define a + b pointwise. That is: (a + b) i = a i + b i for all i. We say a b if a + b T. Then we define a b = a + b. We define an equivalence relation on T as follows: a b if there is a c N n such that both a + c, b + c T (E). Theorem 35 ([4]). Given a finite effect algebra E such that 0 1. Then: E is isomorphic to T /,, [(0,..., 0)], T. Proof. We would like to prove T /,, [(0,..., 0)], T is an effect algebra and then prove it is isomorphic to E. However, for arbitrary T, it is not an effect algebra. That is why we first prove that there is an operation preserving bijection ϕ and then conclude the latter is an effect algebra and hence ϕ an isomorphism. 16

17 Given b E. Suppose c and c are tuples of natural numbers such that b = c 1 a 1 c n a n = c 1a 1 c na n. Let d be such that Then: b = d 1 a 1 d n a n. 1 = b b = (c 1 + d 1 )a 1 (c n + d n )a n = (c 1 + d 1 )a 1 (c n + d n )a n. Hence: c + d, c + d T. Thus: c c. Define ϕ: E T / by ϕ(b) = [c]. Given c T. Then c t for some t T. b = c 1 a 1 c n a n is defined, since c t and t is a multiplicity vector. By definition: ϕ(b) c. Thus ϕ is surjective. Given b, b E with ϕ(b) = ϕ(b ). Thus: there are tuples c, c and d such that b = c 1 a 1 c n a n ; b = c 1a 1 c na n and b + d, b + d T. Note d t for some t T. Hence d = d 1 a 1 d n a n is defined. Furthermore: d b = d b = 1. By canceling: b = b. Thus ϕ is injective. Given tuples c, c and d such that c c and c d. Then c + d T and: (c 1 + d 1 )a 1 (c n + d n )a n = (c 1 + d 1 )a 1 (c n + d n )a n 1. Hence c d and c + d c + d. Thus and can be extended to T /. Suppose b, b E with b b. Let c and c be such that: b = c 1 a 1 c n a n b = c 1a 1 c na n. Then: b b = (c 1 + c 1)a 1 (c n + c n)a n. Hence c + c ϕ(b b ). Thus c + c T. That is: c c. Also ϕ(b) ϕ(b ) = ϕ(b b ). Conversely, suppose ϕ(b) ϕ(b ). Again, find c and c such that: b = c 1 a 1 c n a n b = c 1a 1 c na n. Then c c. Thus (c 1 + c 1)a 1 (c n + c n)a n is defined. Thus b b. Finally, clearly ϕ(1) = T. The operations on T / are preserved by the surjective ϕ. Hence it is an effect algebra. Furthermore: since ϕ is injective, we have T / = E. 17

18 2.2 Effect monoids Various examples of effect algebras also carry a multiplication. We will consider effect monoids, which are effect algebras with an associative and distributive multiplication. They play an important rôle in one class of effect logics, see Subsection Definition 36. A structure E,,, 0, 1 is called an effect monoid if E,, 0, 1 is an effect algebra and the (total) binary operation satisfies the following. (M1) (unit) a 1 = 1 a = a. (M2) (left distributivity) if a b, then c a c b and (c a) (c b) = c (a b). (M3) (right distributivity) if a b, then a c b c and (a c) (b c) = (a b) c. (M4) (associativity) a (b c) = (a b) c. Example 37. We can extend the first two effect algebras of Example 2 to effect monoids as follows. 1. [0, 1], +,, 0, 1 where is the standard multiplication on R. 2. P(X),,,, X where is the intersection. Proposition 38. Given an effect monoid E, we have 1. a b a and a b b for any a, b E; 2. if a b then c a c b and a c b c; 3. if a b then c a c b and a c b c; 4. a b = a (a b) and a b = b (a b) and 5. whenever c b, we have a (b c) = (a b) (a c). Proof. One by one. 1. Certainly b b. Thus a b a b and (a b) (a b ) = a (b b ) = a 1 = a. Hence a b a. The argument for the other statement is similar. 2. Suppose a b. Then a d = b for some d. Hence (c a) (c d) = c (a d) = c b. Consequently c a c b. The argument for the other statement is similar. 3. For every n N we have na b. Hence by the previous n(c a) = c na c b. Thus c a c b. The proof of the other statement is similar. 4. Consider that b b and thus (a b ) (a b) = a (b b ) = a. By uniqueness of the difference, we know a b = a (a b). The other proof is similar. 18

19 5. If c b, then b c is defined and b c c. Hence (a (b c)) (a c) = a ((b c) c) = a b and thus by uniqueness of the difference we have a (b c) = (a b) (a c). Definition 39. An effect monoid E is called commutative if for every a, b E we have a b = b a. The previous two examples are both commutative. We will not find a finite effect monoid that is non-commutative. Proposition 40 ( ). If E is a finite effect monoid, then there exists a finite set X such that E = P(X),,,, X. Proof. Let a 1,..., a n be the atoms of E. If i j, then a i a j a i, a j hence a i a j = 0. Also note that a i a i = 0 or a i a i = a i. Given a multiplicity vector t 1,..., t n, that is t 1 a 1 t n a n = 1, then a j = 1 a j = (t 1 a i t n a n ) a j = t j (a j a j ) a j. Thus a j a j = a j and consequently t j = 1 for any j. Hence: the only multiplicity vector of E is (1,..., 1). Furthermore, given any b E, we know b = b 1 a 1 b n a n with b i {0, 1}. Thus b b = b b. Pick any X with X = n. Then P(X),,,, X also has n atoms and exactly one multiplicity vector: (1,..., 1). Thus by Theorem 35, it is, as an effect algebra, isomorphic to E. Also a b = a b. Thus it is, as an effect monoid, isomorphic to E. The effect monoid structure on [0, 1] is unique. Proposition 41. If makes [0, 1] R an effect monoid, then is the standard multiplication on R. Proof. First note that n 1 n = 1 and thus x 1 n = x n. Hence x m n = m n x. Given any x, y R. Let q 1, q 2,... Q such that y... q 2 q 1 1 and q i y. Then x y... xq 2 xq 1. And thus x y xy. Approximating y from the other side, we get x y xy. Consequently x y = xy. The crux of the previous proof was to show that has to respect the scalar multiplication on the underlying effect algebra. Because we will encounter such effect monoids again, we define: Definition 42. An effect monoid E is called convex if the underlying effect algebra is convex and λ (a b) = (λ a) b = a (λ b) Convex effect monoids and OAU-algebras We will prove a representation result for convex effect monoids, which we will use to simplify the study of a certain class of effect monoids. Recall that given any ordered vector space V and vector u > 0, we know that the order interval [0, u] is a convex effect algebra (Proposition 20). Conversely, any convex effect algebra is an interval effect algebra of some ordered vector space (Theorem 23). Given any convex effect monoid E. Then in particular it is a convex effect algebra. And thus it is an interval effect algebra of some ordered vector space V. 19

20 We will show that we can extend the multiplication of the effect algebra to the whole vector space, which will form an ordered, associative and unitary algebra. Conversely, given any ordered, associative and unitary algebra with unit 1, we can restrict the algebra multiplication to the order interval [0, 1], which will form an effect monoid. Definition 43. A structure V, +,,,, 0, 1 is called an ordered associative unitary algebra (OAU-algebra) if V, +,,, 0 is a vector space and is a binary operation that satisfies 1. (associativity) a (b c) = (a b) c; 2. (distributivity) (b + c) a = b a + c a and a (b + c) = a b + a c; 3. (unit) 1 a = a 1 = a; 4. ( preserves order) if c 0 and a b, then c a c b and a c b c and 5. (homogeneity) r (a b) = (r a) b = a (r b). Proposition 44. Given an OAU-algebra V. For a, b [0, 1] with a + b 1, define a b = a + b. Then E = [0, 1],,, 0, 1 is a convex effect monoid. Proof. Given a, b [0, 1]. Then 0 a and thus 0 = 0 (0 b) = 0 b a b. Also a 1 and thus a b 1 b = b 1. Thus a b is a total binary operation on [0, 1]. By Proposition 20, E is a convex effect algebra. Now, we check the convex effect monoid axioms. (M1), (M4) and convexity are immediate from the OAU-algebra axioms. To prove (M2), assume a, b [0, 1] with a + b 1. Then c a + c b = c (a + b) c 1 = c 1, as desired. The proof of (M3) similar. Before we prove that any convex interval effect monoid can be extended to an OAU-algebra, we need a small result on ordered vector spaces. Lemma 45 ( ). Given an ordered vectorspace V and an element u > 0. The following are equivalent. 1. [0, u] generates V, see Definition Given v V, we have v = r(v 1 v 2 ) for some v 1, v 2 [0, u] and r [1, ). 3. u is a strong unit, that is: for every v V, there is a n N such that v nu. Proof. 1. First we prove (i) implies (ii). Thus suppose [0, u] generates V. Given v = r 1 v 1 r 2 v 2. Suppose r 1, r 2 < 0. Then v = r 2 v 2 r 1 v 1. Thus we may assume not both: r 1, r 2 < 0. Suppose only r 1 < 0. Note 1 2 (v 1 + v 2 ) [0, u]. Thus 2(r 2 r 1 ) 1 2 (v 1 + v 2 ) 0 0 = r 1 v 1 r 2 v 2. Note r 2 r 1 0. Thus we may assume r 1 0. Similarly, we may assume r 2 0. Suppose both r 1, r 2 [0, 1]. Then r 1 v 1, r 2 v 2 [0, u] and hence v = 1(r 1 v 1 r 2 v 2 ), which is of the right form. Suppose r 1 [0, 1] and r 2 (1, ). Then r1 r 2 v 1 [0, u] and v = r 2 ( r1 r 2 v 1 v 2 ), thus we are done. We are also done if r 1 (1, ) and r 2 [0, 1]. Finally, suppose r 1, r 2 (1, ). Then 1 r 2 v 1, 1 r 1 v 2 [0, u] and v = r 1 r 2 ( 1 r 2 v 1 1 r 1 v 2 ), which is of the desired form. 20

21 2. Now we prove (ii) implies (iii). Thus suppose (ii). Given v V. Then v = r(v 1 v 2 ) for some r [1, ) and 0 v 1, v 2 u. Note v 1 v 2 u. Hence v = r(v 1 v 2 ) ru r u as desired. 3. Finally, we prove (iii) implies (i). Thus suppose u is a strong unit. In particular, for some n 1, we have nu v nu. Thus: ( u + v n ) u and ( v ) u u. n Now define v 1 = 1 2 (u + v n ); and v 2 = 1 2 (u v n ). We have v = nv 1 nv 2 as desired. Theorem 46 ( ). Given an ordered vector space V and a vector u > 0 such that [0, u] generates V. Suppose [0, u],,,, 0, u is a convex effect monoid. Then there is a unique extension of to V such that V, +,,,, 0, u is an OAU-algebra. Proof. Suppose is an extension of to V such that V, +,,,, 0, u is an OAU-algebra. Write U = [0, u]. Given a, a V. Then a = r(v w) and a = r (v w ) for some r, r (1, ) and v, v, w, w U. Hence a a = r(v w) r (v w ) = rr (v v + w w w v v w ) = rr (v v + w w w v v w ). (1) Thus the extension, if it exists, is unique. It also suggests a definition for a a. However, we need to show that the choice of r, r, v, v, w and w does not effect the value of (1). We do this in two steps. First we define on U V. Then extend it to V V. Given r, r [1, ) and x, v, v, w, w U. Without loss of generality, we may assume r r. Suppose r(v w) = r (v w ). We want to show that r(x v x w) = r (x v x w ). From the assumption and thus by dividing by 2r gives Note that r r [0, 1] and thus r r w, r rv + r w = r v + rw 1 2 v + r 2r w = r 2r v w. also 1 2 u u U. Thus 1 2 v + r 2r w = r 2r v r 2 (x v) + 2r (x w ) = x ( 1 2 v + r 2r w ) = x ( r 2r v w) r v U. Furthermore, if u, u U, then w U. And consequently = r 2r (x v ) (x w). Rearranging and multiplying by 2r, yields the desired r(x v x w) = r (x v x w ). 21

22 And thus we can define x a = r(x v x w) if a = r(v w) for r [1, ) and x, v, w U. We want to repeat this argument to define x y for x, y V, by x y = r(v y w y). If we review the argument, we see we need to check whether (a + b) y = a y + b y and s(a y) = (sa) y for s [0, 1] and a, b, a + b U. We check the latter first. Suppose s [0, 1], a U and y V with y = r(u v) for some r [0, 1] and u, v U. Then s(a y) = sr(a v a w) = r(a (sv) a (sw)) = a (sy). Now, for the partial distributivity, additionally assume a, b, a + b U. Then (a + b) y = r((a + b) v (a + b) w) = r(a v + b v (a w + b w)) = r(a v a w) + r(b v b w) = a y + b y. Thus we can indeed repeat to previous argument and define x y = r(v y w y), which is the same as (1). Finally, we need to check whether obeys the axioms of an OAU-algebra. We do this in a convenient order. Suppose a, b, c V ; a = r a (v a w a ); b = r b (v b w b ) and c = r c (v c w c ). 1. (distributivity) Assume, without loss of generality that r b r c. Then r b r c ( [0, 1] and we have a + b = 2r c ( r b 2r c v b v c) ( r b 2r c w b w c) ). Hence ( a (b + c) = 2r a r c va ( r b 2r c v b v c) + w a ( r b 2r c w b w c) v a ( r b 2r c w b w c) w a ( r b 2r c v b v c) ) ( = 2r a r rb c 2r c (v a v b ) (v a v c ) + r b 2r c (w a w b ) (w a w c ) r b 2r c (v a w b ) 1 2 (v a w c ) r b 2r c (w a v b ) 1 2 (w a v c ) ) = r a r b (v a v b + w a w b v a w b w a v b ) + r a r c (v a v c + w a w c v a w c w a v c ) = a b + a c. The argument for right distributivity is similar. 2. (homogeneity) Suppose r R. We distinguish cases. If r [1, ), then ra = rr a (v a w a ) with rr a [1, ) and thus (ra) b = rr a r b (v a v b + w a w b v a w b w a v b ) = r(a b). Suppose r [0, 1]. Then ra = r a (rv a rw a ) with rv a, rw a U and thus (ra) b) = r a r b ((rv a ) v b + (rw a ) w b (rv a ) w b (rw a ) v b ) = r a r b r(v a v b + w a w b v a w b w a v b ) = r(a b). 22

23 Suppose r = 1. Then a = r a (v a w a ) = r a (w a v a ). And thus ( a) b = r a r b (w a v b + v a w b v a v b w a w b ) = r a r b (v a v b + w a w b v a w b w a v b ) = (a b) = r(a b). For the remaining case, suppose r < 0. We can reduce it to the previous cases: (ra) b = ( r ( a)) b = (( ra) b) = ( r)(a b) = r(a b). The argument for r(a b) = a (rb) is similar. 3. (associativity) Using the homogeneity and distributivity we just demonstrated, we can reduce the associativity of to that of, as follows. a (b c) = a (r b r c (v b v c + w b w c v b w c w b v c )) = r b r c ( a (vb v c ) + a (w b w c ) a (v b w c ) a (w b v c ) ) = r b r c ( (ra (v a w a )) (v b v c ) + (r a (v a w a )) (w b w c ) (r a (v a w a )) (v b w c ) (r a (v a w a )) (w b v c ) ) = r a r b r c ( va (v b v c ) + v a (w b w c ) + w a (v b w c ) + w a (w b v c ) v a (v b w c ) + v a (w b v c ) w a (v b v c ) + w a (w b w c ) ) = r a r b r c ( (va v b ) v c + (v a w b ) w c + (w a v b ) w c + (w a w b ) v c (v a v b ) w c + (v a w b ) v c (w a v b ) v c + (w a w b ) w c ) = r a r b ( (va v b ) c + (w a w b ) c (v a w b ) c (w a v b ) c ) = r a r b (v a v b + w a w b v a w b w a v b ) c = (a b) c 4. (unit) u a = r a (u v a u w a ) = r a (v a w a ) = a. Similarly a u = a. 5. ( preserves order) First note that if a 0, then r a (v a w a ) 0 and thus v a w a 0. Also v a w a u, thus v a w a U. Thus a = rv for some v U viz v = v a w a. Next, suppose c, a 0. With the previous we may assume c = r c v c and a = r a v a. Thus c a = (r c v c ) (r a v a ) = r c r a (v c v a ) = r c r a (v c v a ). We know v c v a U. In particular v c v a 0. Also r c r a 0. Thus c a 0. Finally, suppose c 0 and a b. Then b a 0. Hence 0 c (b a) = c b c a. Thus c a c b, as desired. The other case is similar Effect monoids on finite dimensional lexicographically ordered vector spaces We want to study effect monoids that are not commutative. Suppose we find a non-commutative OAU-algebra. By Proposition 44, its unit interval is an effect monoid. It is not hard to see it must be non-commutative too. 23

24 In this section we will study the class of effect monoids derived from OAUalgebras on lexicographically ordered vector spaces. This will give us examples of non-commutative effect monoids. For this section, assume n N and n 1. We write e 1,..., e n for the standard basis of the real vector space R n. Given a vector v R n, we assume v 1,..., v n R are the components; that is: i v ie i = v. We can totally order R n as if it were words in a dictionary. Definition 47. Given n N. Given v, w R n, we say v < w if there is an i such that v i < w i and for all j < i we have v j = w j. R n with this order is an ordered vector space, which is called lexicographically ordered. Since the order is total, we can familiarly define { v v 0 v = v v 0. We write v w if for all n N we have n v w. Note that 0 e n e n 1 e 2 e 1. The order interval [0, e 1 ] generates R n and thus [0, e 1 ] is a convex effect monoid. See Section Call it Elex n. We are interested in the effect monoids on En lex. First, we will prove that any effect monoid on Elex n is convex. Then by Theorem 46 we know that an effect monoid on Elex n extends uniquely to an OAUalgebra on R n. We will show that an OAU-algebra on R n is fixed by e i e j. Then we will give necessary and sufficient conditions to extend a multiplication defined on the standard basis to an OAU-algebra. Lemma 48. Any effect monoid on Elex n (the unit interval effect algebra of the n- dimensional lexicographically ordered vector space) is convex. Proof. Given n N and a, b Elex n. Certainly n(a 1 n b) = a n( 1 n b) = a b and thus a ( 1 n b) = 1 n (a b). Then also for any 0 m n, we have m n (a b) = a ( m n ) b. Similarly m n (a b) = ( m n a) b. For any r [0, 1] we can find q i, q i Q [0, 1] such that r is the uniquely defined by q i r q i for all i. Suppose a b = 0. Then certainly a (r b) a b = 0 = r(a b). Suppose a b 0. Then a b > 0 and if q i (a b) r (a b), then q i r. Thus r (a b) is uniquely defined by q i (a b) r (a b) q i (a b) for all i. Now note that q i (a b) = a q i b a (r b) a q i b = q i (a b). Thus a (r b) = r (a b). Similarly (r a) b = r (a b). Given any OAU-algebra on the lexicographically ordered R n. Note that by the homogeneity and distributivity of, that is: its bilinearity, we have v w = ( ) ( ) v i e i v j e j = v i v j (e i e j ). i j Write e ij = e i e j. We see is fixed by the vectors e ij. Conversely, given vectors e ij, we can define a multiplication by i,j v w = i,j v i v j e ij. 24

25 However, this does not necessarily form an OAU-algebra. necessary and sufficient conditions. The following are Proposition 49. Given vectors e ij. Write v w = i,j v iw j e ij. The following are equivalent. R n, +,,,, 0, e 1 is an OAU-algebra. The following four conditions hold. e 1j = e j and e i1 = e i e i (e j e k ) = (e i e j ) e k e ij 0 If i < j, then e ik e jk and e ki e kj. Proof. The necessity is clear. To prove sufficiency of the four conditions, we check the axioms of an OAU-algebra in a convenient order. 1. (distributivity) Given a, b, c R n. Then a (b + c) = i,j a i (b j + c j )e ij = i,j a i b j e ij + i,j a i c j e ij = a b + a c. Right distributivity is proven similarly. 2. (homogeneity) Given a, b R n and r R. Then r(a b) = r i,j a ib j e ij = i,j (ra i)b j e ij = (ra) b. The other case is proven in the same way. 3. (associativity) Given a, b, c R n. By the second condition and the distributivity and homogeneity just proven, we have a (b c) = a j,k b j c k e jk = j,k b j c k (a e jk ) = j,k ( ) b j c k a i e i (ej e k ) i = i,j,k a i b j c k (e i (e j e k )) = i,j,k a i b j c k ((e i e j ) e k ) = (a b) c. 4. (unit) From the first condition and the definition of we get e 1 a = j a je 1j = j a je j = a and similarly a e 1 = i a ie i1 = i a ie i = a. 25

26 5. ( preserves order) If we can prove that a b 0, whenever a, b 0, then we are done. For suppose c 0 and a b. Then b a 0. Thus 0 c (b a) = c b c a. And thus c a c b, as desired. The other case is similar. Thus, suppose a, b 0. If a = 0 or b = 0, then a b = 0 0. Thus, we may assume a, b > 0. Then there are α and β such that a α, b β > 0 and a i = 0 for all i < α and b i = 0 for all i < β. Consider a α b β e αβ. By the third condition and the current assumptions, we know a α b β e αβ 0. Given i and j with (i, j) (α, β). We will show a α b β e αβ a i b j e ij, by distinguishing cases. Suppose i < α. Then a i = 0 and thus a i b j e ij = 0 a α b β e αβ. The same argument covers the case j < β. Suppose i > α and j β. By the fourth condition, we know e αβ e iβ e ij. And since a α b β > 0, we also have a α b β e αβ a i b j e ij. The case i α and j > β is similar. Note that if 0 v then for any w v we have 0 v + w. And thus a b = a i b j e ij = a α b β e αβ + a i b j e ij 0. i,j (i,j) (α,β) Corollary 50 ( ). The unique effect monoid on E 2 lex is: e 1 e 2 e 1 e 1 e 2 e 2 e 2 0 Corollary 51. There is a non-commutative effect monoid on Elex 5, fixed by: e 1 e 2 e 3 e 4 e 5 e 1 e 1 e 2 e 3 e 4 e 5 e 2 e 2 e 4 e e 3 e e 4 e e 5 e

27 2.3 Effect modules Recall that a convex effect algebra is an effect algebra equipped with a scalar multiplication with [0, 1]. See Definition Effect modules are a generalization of convex effect algebras, where the scalars can come from any effect monoid. Definition 52. Given an effect monoid M. An M-effect module is an effect algebra E together with an operation ( ) ( ): M E E such that (V1) α (β a) = (α β) a; (V2) if α β then αa βa and (α β) a = α a β a; (V3) if a b then λ a λ b and λ a λ b = λ (a b) and (V4) 1 a = a. Example Every convex effect algebra is a [0, 1]-effect module. 2. Every effect algebra is a 2-effect module with 0 a = 0 and 1 a = a. 3. Given any effect monoid M and n N, the set M n is a M-effect module with pointwise operations. 4. A bit more general: given any effect monoid M and set X, the set M X of functions from X to M is an M-effect module with pointwise operations. Definition 54. Given a map between M-effect modules f : E 1 E 2. f is an effect module homomorphism if f is an effect algebra homomorphism and furthermore f(λ a) = λ f(a) for all λ M and a E. We write EMod M for the category of M-effect modules with effect module homomorphisms. 27

28 2.4 Sequential effect modules Recall that the starting point of this thesis, was the observation that in the examples of effect logics initially studied by Jacobs, a sequential effect algebra arises. We did not define this notion, yet. See Subsection 1.1. Definition 55 ([7]). A sequential effect algebra is an effect algebra E together with a binary multiplication such that (S1) a (b c) = (a b) (a c) (S2) 1 a = a (S3) If a b = 0, then a b = b a. (S4) If a b = b a, then a b = b a and a (b c) = (a b) c. (S5) If c a = a c and c b = b c. Then: c (a b) = (a b) c and c (a b) = (a b) c. Definition 56. A sequential effect module is a sequential effect algebra, where the underlying effect algebra is an effect module and (SM) λ(a b) = (λa) b = a (λb) for any scalar λ. Definition 57. A sequential effect algebra E is called commutative if for any a, b E, we have a b = b a Examples We have seen commutative sequential effect algebras already, in disguise. Proposition 58. Every commutative effect monoid is a commutative sequential effect algebra. And, conversely, every commutative sequential effect algebra is a commutative effect monoid. Furthermore, the commutative effect monoid is convex if and only if the commutative sequential effect algebra is a [0, 1]-effect module. Proof. 1. Given a commutative effect monoid E. The axioms (S1) and (S2) are satisfied directly by definition. The axioms (S3), (S4) and (S5) are implications of which the conclusions are directly satisfied by definition. If the effect monoid is convex, then (SM) follows by definition. 2. Conversely, given a commutative sequential effect algebra. (M3) is the same as (S1). Since everything commutes, (M3) implies (M2) and (S4) implies the associativity of ; that is: (M4). We are left to prove (M1). By (S1), we have (a 0) (a 1) = a 1. By cancellation: a 0 = 0. Thus by (S3), we have 0 a = a 0 = 0 and by (S4) and (S2) also 1 a = a 1 = a. That is: we have shown (M1). If the sequential effect algebra is a sequential [0, 1]-effect monoid, then the underlying effect algebra is convex and the multiplication is bi-homogeneous, hence (SM). Now, the prime example of a sequential effect module: 28